Cold Storage

Every company has a graveyard of data it legally cannot delete but practically never reads. Healthcare records from seven years ago. Financial transaction logs from a closed fiscal year. Server logs from a service that was decommissioned in 2021.

This data sits on the same storage as your active production data. Same SSDs. Same IOPS. Same monthly bill. And that's absurd, it's like paying Manhattan rent for a storage unit full of boxes you open once a decade.

Cold storage exists to solve this economic mismatch. It's storage optimized for one thing: keeping data alive at the lowest possible cost. You sacrifice fast retrieval in exchange for prices that are 10-25x cheaper than standard storage.

AWS Glacier, for instance, charges $0.004 per GB per month. Compare that to $0.023 for S3 Standard. At 100TB, that's the difference between $400/month and $2,300/month. Over a year, you save $22,800, and that's just one storage tier for one dataset.

Cold storage is defined by three characteristics:

1. Optimized for writes, not reads. Data goes in easily. Getting it back out takes time, minutes, hours, or even half a day depending on the tier.

2. Cheap per-gigabyte, expensive per-retrieval. The storage cost is rock-bottom, but you pay a premium every time you actually need to read something. This pricing model only makes sense if you rarely retrieve data.

3. High durability, low availability. Cold storage systems are designed to never lose your data (99.999999999% durability is common). But "available to read right now" is not guaranteed, retrieval requires a request that may take time to fulfill.